[OpenAFS-devel] semantics of stat'ing a mount point

[Derrick J Brashear]

In previous AFS versions, stat'ing an AFS mount point returns ENODEV for
an unknown volume, but ENOENT if the mount point is for an unknown cell.

A patch proposes to change that to consistently return ENODEV. Is anyone
relying on the distinction? Cases could be made for
-ENOENT represents a missing entry in the CellServDB for the cell
-mount points should be treated like dangling symlinks and always return
ENOENT when invalid

The latter would be a considerable semantic change, though.

-D